Table of Contents

Mobile application users behavior analysis

Atachments

Research purpose

Research description

Analyze the impact of events on the completion of the target event — viewing contacts

The dataset contains data on events committed in the "Unnecessary Things" mobile application. In it, users sell their unwanted items by posting them on a bulletin board.


The dataset contains users data that was performed in the application for the first time after October 7, 2019.

Columns in mobile_sources.csv:

Columns in mobile_dataset.csv:

Types of actions:

Data preprocessing

Columns names correction

Data and time columns creation

Interim summary

There are no Duplicates nor gaps in both datasets.
Column names has been brought to unified style.
Observation period is nearly a month: 2019-10-07 - 2019-11-03

Data analysis

Measuring main parameters (unique users per action etc)

Let's interchange 7 events "show_contacts" to "contacts_show" as they seem literally single-valued.

Division by sessions

The logic behind division by session has been taken as follows:
We assume that session is a flow of events taken by unique user with timediffer between events of one user more then 20 minutes.

Anomaly detection and elimination

As our goal is to analyze the impact of preceding events on the target event "contacts_show",
we can eliminate sessions consisting of 2 unique events by creating following filters by sessions:

Creation of a new dataset "data2" containing only sessions longer than 1 event and with variety of 2 or more events

Interim summary

10975 sessions has been allocated.
New dataset 'data2' has been created: containing only sessions longer than 2 unique events and containing target event. In result:
Data has been reduced to 29492 from 74197 and
Number of sessions has been reduced to 2296 from 10975

Main purpose of the study

Histogram of events

After clearing the data it appeared that the most popular events "tips_show", "photos_show" and "search_1" in 70% of the time is involved into pointless sessions conisting only of the one event.
Directed conscious actions demanding an active action from user like "advert_open", "contacts_show" and "map" has shown less reduction.

Division by events chains

Impact of events leading to target event by all funnel scenarios

From Funnel we can see that the most common events leading to "Contacts show" are the following:


We can see that the user action "tips_show" and "map" are the most common ways to reach target event.
But this actions make only around 25% conversion to "Contacts show".

It seems that the raise of conversion of "tips_show" and "map" will increase number of "Contacts show" events.

"Photos_show" and "search_1" make around 67% conversion to "Contacts show".

Raise of commiting "photos_show" and "search_1" is also an option, as they show the biggest conversion.

"Favorites_add" makes 42% conversion to "Contacts show".

Raise of commiting "Favorites_add"

Also "tips_click" is showing pretty good conversion -38%, but takes a small part of events flows.

Raise of commiting "tips_click"

"Advert_open" makes the smallest conversion to target event - 15%

Raise of conversion "Advert_open"

Timediffer between the prevalent events

Interim summary

We've substracted 5 most popular sessions scenarios:
advert_open, map, tips_show 300
contacts_show, map, tips_show 176
advert_open, map, search_3, tips_show 86
search_4, search_5, search_6, tips_show 80
contacts_show, photos_show, search_1 75

We can see that the user action "tips_show" and "map" are the most common ways to reach target event.
But this actions make only around 25% conversion to "Contacts show".

It seems that the raise of conversion of "tips_show" and "map" will increase number of "Contacts show" events.

"Photos_show" and "search_1" make around 67% conversion to "Contacts show".

Raise of commiting "photos_show" and "search_1" is also an option, as they show the biggest conversion.

"Favorites_add" makes 42% conversion to "Contacts show".

Raise of commiting "Favorites_add"

Also "tips_click" is showing pretty good conversion -38%, but takes a small part of events flows.

Raise of commiting "tips_click"

"Advert_open" makes the smallest conversion to target event - 15%

Raise of conversion "Advert_open"

Mean timediffer between prevalent events and target "Contacts show" is around 300 seconds (5 minutes)

Hypotheses statement

First hypothesis check

H0: Conversion to contact views is equal between two groups: ones who perform tips_show and tips_click actions, and ones with only tips_show.
H1: Conversion to contact views between two groups differs

Second hypothesis check

H0: Conversion to contact views is equal between group of users who performed favorites_add and who did not.
H1: Conversion to contact views between two groups differs

Third hypothesis check

H0: Conversion to contact views is equal between group of users who used photos_show and who did not.
H1: Conversion to contact views between two groups differs

Interim summary

There are no statistically significant differences in conversion to "contacts_show" between the ones who perform
tips_show and tips_click actions, and ones with only tips_show .

There are statistically significant differences in conversion to "contacts_show" between the users who performed
favorites_add and who did not.

There are no statistically significant differences in conversion to "contacts_show" between the users who performed
photos_show and who did not.

Project summary

10975 sessions has been allocated and has been reduced to 2296 sessions after eliminating double unique event sessions.


The most common scenarios of successful event flow are:


"Tips_show" and "map" are the most common ways to reach target event.
But this actions make only around 25% conversion to "Contacts show".

"Photos_show" and "search_1" make around 67% conversion to "Contacts show".

"Favorites_add" makes 42% conversion to "Contacts show".

Also "tips_click" is showing pretty good conversion -38%, but takes a small part of events flows.

"Advert_open" makes the smallest conversion to target event - 15%


Mean timediffer between prevalent events and target "Contacts show" is around 5 minutes


There are statistically significant differences in conversion to "contacts_show" between the users who performed
favorites_add and who did not.

There are no statistically significant differences in conversion to "contacts_show" between the ones who perform
tips_show and tips_click actions, and ones with only tips_show .

There are no statistically significant differences in conversion to "contacts_show" between the users who performed
photos_show and who did not.

Recommendations

In the order of impact:

  1. Raise the conversion of "tips_show" and "map"
  2. Increase commission "photos_show" and "search_1"
  3. Increase commission "Favorites_add"
  4. Increase commission "tips_click"
  5. Raise the conversion "Advert_open"